THE INFLUENCE OF MASS AND ELITE FOREIGN POLICY BELIEFS ON JUDGMENTS OF PRESIDENTIAL PERFORMANCE

Transcription

1 THE INFLUENCE OF MASS AND ELITE FOREIGN POLICY BELIEFS ON JUDGMENTS OF PRESIDENTIAL PERFORMANCE Richard W. Boyd and Jeffrey B. Lewis Department of Government Wesleyan University Middletown, CT November, 1991

2 ABSTRACT This analysis supports the theory that there is a hierarchical structure to the belief systems of both the general public and opinion leaders. People do appear to behave as cognitive misers. Even if they have limited direct experience with international affairs, they can identify general goals for U.S. foreign policy, and they can link these goals to more specific policies. Similarly, people also seem to base their judgments of presidential performance in foreign affairs on the policy positions they hold. A social cognition theory of belief systems is supported in this analysis. The data bases are the cross-section and leadership surveys conducted for the Chicago Council on Foreign Relations in The comparison of the cross-section and leadership samples indicates that the belief systems of ordinary citizens are as hierarchically structured as the beliefs of leaders. One major difference between leaders and non-leaders is that leaders more clearly distinguish human rights values from the goals of protecting American economic and military security. They appear less inclined to believe that American interests are consonant with the interests of other nations. A second important difference is that elites differentiate the goals of military and economic security more sharply than do ordinary citizens. This is a revised version of a paper presented at the Joint Annual Convention of the British International Studies Association and the International Studies Association London, March 28-April 1, 1989

3 To what extent and by what cognitive process do ordinary citizens come to political judgments on foreign policy issues? Aldrich, Sullivan, and Borgida (1989) begin their analysis of foreign policy issues in U.S. elections by noting a paradox: American presidential candidates campaign as though they believe that foreign issues are central to elections. Yet, the literature on voting decisions has until recently offered weak support for this belief because most researchers have concluded that domestic issues dominate most elections. We argue that the literature on public opinion and foreign policy underestimates the degree to which issues of foreign affairs influence people's political evaluations. Foreign policy issues were significant factors in vote choices in the post-war elections of 1952, 1972, 1980, and 1984 (Aldrich, Sullivan, and Borgida, 1989, and Boyd, 1986 and 1988). Here, we investigate the belief systems that lie behind evaluations of presidential performance. People do have general foreign policy goals. These goals influence their policy preferences, which in turn determine their judgments of presidential performance. To the degree that such evaluations affect vote choice, there is a measure of democratic accountability in foreign policy. A key to this argument is a reinterpretation of belief systems in terms of social cognition theory. Social Cognition Theories An extensive literature supports the view that issues are fundamental to most people's vote choices (e.g., RePass, 1971; Boyd, 1972; Pomper, 1972; Nie, Verba, and Petrocik, 1976; Asher, 1984). These studies serve several important functions. They provide empirical support for the

4 rational choice, spatial models of Downs (1957) and his successors (e.g., Jeffrey Smith, 1980 and Enelow, Hinich, and Mendell, 1986). They effectively challenge arguments that there are immutable cognitive limits to the capacity of citizens to bring informed preferences to bear upon political decisions. And, they underline the degree to which informed choice rests upon opinion leadership, a context in which candidates and other party leaders cultivate public support with issue-based appeals. Even in the arena of foreign affairs, more recent research indicates that foreign issues are very salient (Tom Smith, 1980) and no more likely than domestic issues to be subject to rapid fluctuations of sentiment (Page and Shapiro, 1982). What the issue-voting literature did not do, however, was to ground its investigations in sophisticated theories of cognitive psychology. Fortunately, a line of research now integrates rational-choice and cognitive psychology theories of political judgments. Social cognition theory focuses on the ways in which people process information about a complex world. (See Fiske and Taylor, 1984, and Conover and Feldman, 1984, for a review of this perspective.) People who must make judgments on matters so complex as to be beyond full understanding employ what are variously termed information short-cuts, cognitive heuristics, or "mental economies" (Gant and Davis, 1984). Being "cognitive misers," people process new and specific information by interpreting it in the light of their more general perspectives or ideals. When a new issue such as aid to the Nicaragauan Contras in the mid-1980s becomes a public controversy, people try to interpret the information by resort to general values they already hold, such as the virtue of defending allies (however identified) or a belief that Communism is a source of revolutionary movements. 2

5 This mode of processing information is sometimes called schematic reasoning. That is, people interpret new information by resort to schemas (sometimes called stereotypes or general postures). These schemas are available to the ordinary voter because they are part of the common culture. There are many potential schemas in which any new information can be presented. And, of course, the media and political leaders can influence people's cognitions by suggesting, selecting, manipulating or "framing" the schemas through which information is presented (Iyengar, 1987). A central element of this view of cognition is that "domain-specific information is structured in memory in cohesive kernels of thought, and it is this 'unitized' body of knowledge that makes information meaningful" (Hamill, Lodge, and Blake, 1985). That is, factual knowledge is structured by "associational knowledge," the cognitive connections that form the network of a person's related beliefs. The conception is similar to what Downs called "contextual knowledge" (Downs, 1957). Jon Hurwitz and Mark Peffley have written a series of important contributions to the social cognition perspective on belief systems, with special attention to foreign policy attitudes (Peffley and Hurwitz, 1985; and Hurwitz and Peffley, 1987a, 1987b, 1988, and 1990). Their article, "How are Foreign Policy Attitudes Structured" (1987a), presents a hierarchical model of attitude organization. They argue that people reason from the most abstract "core values" (morality of warfare and ethnocentrism) down to "general postures" (militarism, anti-communism, and isolationism) down to specific issues, such as support for defense spending. The associational links, then, are vertical and connect attitudes at different levels of abstraction. In their theory 3

6 there are no direct, associational links between attitudes at the same level of abstraction. These core values and general postures are important elements in cognition because people can hold these values even in the absence of direct experience with or substantial information about the attitude object. We hypothesize that in the absence of truly discontinuous changes in the world order, general values are usually stable and are altered by experience only slowly over time. Holsti and Rosenau (1984) offer evidence that elites did gradually change their assumptions about containment and the cold war in the Vietnam era. Rapid and substantial value change, however, occurs only in a context of truly dramatic discontinuities. Recent events such as the disbanding of the Warsaw pact and the weakening of the Soviet Union may represent two such shocks. However, for the period of this study, where no such discontinuous events took place, we assume that general values are stable and that they are not influenced by beliefs on particular issues. Wittkopf (1990) and Chittick, Billingsley, and Travis (1990) support this assumption about the stability of general values toward internationalism. One implication of the staability of general values is that we can assume that there is no reciprocal causation between these values and more specific attitudes. Hurwitz and Peffley (1988a) conducted a panel study in which they tested the reciprocal influences of general values and specific attitudes on each other. They found that the impact of particular attitudes on general values to be small compared to the influence of values on attitudes. This is the basis for our assumption that, within the scope of this study, values influence more specific policy attitudes, and not vice versa. 4

7 The novelty and the utility of the social cognition model is the characterization of these values that lie at the causal origin of a cognitive structure. The earlier theoretical tradition tended to conceive of ideology as the most general and integrative attitudes organizing a belief system. Typologies of "levels of sophistication" (Campbell, 1960, pp. 220ff), however, found fewer than 10 percent of citizens capable of interpreting politics in abstract, ideological terms. In contrast, the social cognition approach sees overarching attitudes as easily accessible to ordinary citizens. These core values and general postures are the language in which political leaders and the media speak to people. In this respect, these general values or schemas have the attributes that Carmines and Stimson (1989) claim for "easy issues." These kinds of issues frame the political discourse of ordinary citizens. A Hierarchical Model of Foreign Policy Judgments Taylor and Crocker (1981, p. 92) argue that all schemas tend to share a "pyramidal structure," in which more specific information is organized by or nested within more abstract beliefs or values. (See also Conover and Feldman, 1984, p. 97). Figure 1 presents a pyramidal model of the structure of foreign policy attitudes, following the theoretical lead of Hurwitz and Peffley. 1 1 The curved lines represent correlations between exogenous variables that are not explained within the model. Such lines are conventionally terminated at both ends with arrows. The straight lines should all be terminated with an arrow pointing downwards, indicating causal connections from general goals to policy positions to performance judgments. 5

8 General Foreign Policy Goals: The model presupposes that people interpret foreign policy events by resort to their general goals for an ideal foreign policy. We assume that most people evaluate policies and events using one or more of three such goals: The first goal is Economic Security the protection of U.S. economic interests, such as the jobs of American workers, a strong dollar, an adequate trade balance, and sufficient energy supplies. The second is Military Security a military posture sufficiently strong to defend the U.S. and American allies abroad. The third is Human Rights. This value represents people's concern with what Wolfers (1962:73ff) termed milieu goals, those more universalistic values not situated in or limited by conceptions of exclusively national interests. These include conditions of life in nations abroad world hunger, standards of living in less developed countries, and human rights and democracy. FIGURE 1 GOES ABOUT HERE Policy Positions: People connect these general goals to particular policies through associational links. Our model includes five broadly defined policy positions: Military Intervention circumstances justifying the use of U.S. troops abroad; Military Assistance including arms sales, aid to friendly authoritarian regimes, and support of covert activities by the CIA; U.S./Soviet Cooperation attitudes on arms control and other bilateral military and economic 6

9 accommodations with the USSR; Economic Assistance to third world nations; and Protectionism attitudes toward protective trade policies for U.S. industries. Evaluations of Presidential Performance in Foreign Affairs: From these policy preferences people make judgments of the administration's performance in managing an inclusive set of foreign policy challenges. This is, of course, a crucial link in a theory of democratic accountability in foreign affairs. This theory of cognitive and evaluative links from general goals to policy preferences to performance judgments carries several important implications. First, like Hurwitz and Peffley we assume that the associational links in each attitude kernel run between levels of abstraction and not within them. That is, most people do not make direct cognitive associations between the policy issues of trade protection and military aid, or between U.S.-Soviet cooperation and military intervention. Rather each policy domain tends to have its own specific links connecting goals, policies, and performance judgments. For this reason we have not hypothesized that all goals are associated in kernels with all policies in what is sometimes called in statistical language a fully saturated model. We have specified only ten associative links between the three goals and the five policy domains. Second, we assume that people do not ordinarily see direct links between their general goals and their performance judgments of a President. Rather, policy preferences form the cognitive bridges between general goals and performance judgments. 7

10 Third, we assume that there are no important, more general beliefs that stand above these general goals. This assumption can surely be challenged. For example, Hurwitz and Peffley (1987a) posit two core values, the morality of warfare and ethnocentrism, as the most general beliefs in a foreign policy attitude structure. Such cultural values are closely related to conceptions of human equality and beliefs about whether human nature is driven primarily by conflict or cooperation. We acknowledge, then, that there may be a level of values above our general foreign policy goals that influence the way in which people make judgments about foreign policy. We have omitted such values because we have no measures of these concepts in our data set. Fourth, we take no position on whether ordinary citizens see associative links between their political predispositions (party and liberal-conservative self-identification) and their general goals for a good foreign policy. We leave the causal origins of the general goals, partisanship, and ideological self-identification unspecified. The curved lines connecting these exogenous concepts (the general goals and political leanings) indicate correlations that are not explained within the model. Fifth, our initial model does not include direct associative links between people's political predispositions and their more specific foreign policy positions. We assume that domestic policies social welfare and race have been the most important determinants of partisanship for most Americans since the New Deal (See Carmines and Stimson, 1989). We include political leanings in our model because the variable obviously has a major, direct influence on people's perceptions of a President's performance in office. Our initial theory is that specific foreign 8

11 policy positions do not mediate the connection between political leanings and performance judgments. Sixth, we hypothesize that ordinary citizens and leaders think about foreign policy issues using the same cognitive schema. This presumption means that we apply the same model specified in Figure 1 to leaders and citizens alike. We assume, for example, that both leaders and mass samples will show evidence of cognitive links between their military security values and their policy predispositions on military intervention, military assistance, and U.S./Soviet cooperation, even if citizens and leaders differ in the strength or direction of these cognitive associations. Wittkopf (1990: ch. 5) and Chittick, Billingsley, and Travis (1990) find significant support for this assumption in their comparisons of the belief systems of citizens and leaders. Data and Measures The data for this test are from the 1986 survey of foreign policy attitudes conducted by the Gallup Organization for the Chicago Council of Foreign Relations. This survey is rich in measures of attitudes toward foreign affairs and includes a cross-section of the American public as well as a separate leadership sample. Appendix A details all of the items that comprise the model. The general goals are measured by fourteen items such as "defending our allies' security" and "promoting and defending human rights in other countries." Respondents are asked to rate the goals on a scale from "very important" to "not at all important." For most Americans all of these values are positive ones. People differ in the degree to which they believe these values represent important national 9

12 aspirations, but few people would appear to regard any of the values as negative. Majorities rated all of the values as either very important or somewhat important in both the cross-section and the leadership samples. These values are examples of what Stokes (1966) terms "valence" rather than "position" items. 2 Our three general values are measured by subsets of the fourteen goal items as described in Appendix B. Our analysis, along with other research, offers evidence that responses to these items reflect three underlying dimensions of foreign policy beliefs. Chittick, Billingsley, and Travis (1989) summarize a number of factor analyses of these items supporting a three-factor solution. Caloss (1989) also finds three dimensions in the patterns of people's responses, although he subsequently forces a two-factor solution to test his particular theory. Our model does differ from those of Wittkopf (1990) and from Chittick, Billingsley, and Travis (1990) in the following important respect: They used factor analyses with orthogonal rotation to construct their latent dimensions (or "faces of internationalism" in Wittkopf's terms). For example, Wittkopf method requires that people's attitudes toward cooperative internationalism be independent of their attitudes toward militant internationalism. Our model presumes that people have common cognitive associations between their basic stances toward military security, economic security, and human rights, and it estimates the strengths of these links. 2 Similarly, Conover and Feldman (1984, p. 108) find that most schemas are organized by positively valenced beliefs. Only one of the foreign policy schemas they isolated was defined by beliefs to which their respondents were in opposition. In effect, then, schemas tend to be distinct perspectives rather than bipolar continua. 10

13 The policy beliefs, unlike the general goals, are more specific, position issues. Responses range from support to opposition. Because valence goals and position issues differ in nature and level of abstraction, we keep them conceptually separate. This permits us to test empirically the relationship between particular goals and policy beliefs. This approach is an important difference between our measures and those of Maggiotto and Wittkopf (1981), Wittkopf (1987), Wittkopf (1990) and Holsti and Rosenau (1990), who factor analyze a broad assortment of items of quite different levels of cognitive specificity to create two underlying general dimensions of Militant and Cooperative Internationalism. 3 The performance judgments of the Reagan Administration are based on twelve items. Eight items measure "the Reagan Administration's handling" of a set of foreign policy areas; four probe evaluations of the performance of the "U.S. government." We test our theory using a covariance structure analysis program, LISREL VI. A major virtue of a covariance structure model is that the survey items are represented as observed indicators of underlying latent dimensions the schemas in our model (Long, 1983a, 1983b). The latent dimensions are modeled in a theory, rather than uncovered by a statistical estimation procedure, such as in ordinary factor analysis. The analyst specifies in advance which items measure particular latent dimensions. The program estimates the loadings of each item on a dimension. As in factor analysis, the constructed, latent dimensions are purged of much of their 3 Holsti and Rosenau (1984) are noted, of course, for creating a three category typology of postwar belief systems: cold war internationalists, post-cold war internationalists, and semiisolationists. However, they have most recently re-conceptualized and remeasured their typology using the Maggiotto-Wittkopf dimensions. See Holsti and Rosenau (1990). 11

14 random error, compared with an alternate procedure of summing observed item responses to create an index. Each observed item contributes to the measure of a latent variable only to the extent that the item is a reliable indicator. This is especially important in the crude measurement world of the typical sample survey. At the same time that LISREL constructs the latent factors, it also estimates any hypothesized causal linkages between the latent factors. The test of a good theory is a measurement and a structural effects model that accurately predict the bivariate associations between all of the observed items. Since many items (61 in the cross-section survey) are used to measure the ten latent variables in Figure 1, missing data is a problem. Listwise deletion of missing data would exclude respondents who fail to answer even one of the items, which would result in a severe sample attrition. We began therefore by omitting the fourteen percent of the cross-section sample who failed to answer more than 25 percent of the items. 4 Then, we used pairwise deletion to create a polychoric correlation input matrix for LISREL. 4 None of the leadership sample exceeds this threshold. 12

15 The Cross-Section Sample Figure 2 presents the correlations among the exogenous (independent) latent variables in the cross-section sample. As predicted for a set of valence attitudes, the general goals are positively intercorrelated in this sample. People who believe strongly in military security also support national economic strength (a correlation of.70), and supporters of military and economic security also tend to value human rights (correlations of about.4). Among ordinary Americans a single nationalist-internationalist dimension appears to influence support for all three general goals. FIGURE 2 GOES ABOUT HERE The index formed from party identification and liberal-conservative self-identification, political leanings, is also related to support for these general goals. The strongest association (.57) links conservatives and Republicans to support for military security, but they also tend to support economic security goals (.26) and to oppose human rights (-.19), compared to Democrats and liberals. Thus, there is a partisan and ideological character to general foreign policy beliefs even in the mass sample, a finding consistent with Wittkopf's analysis (1990, ch. 2). Figure 2 provides evidence of strong cognitive connections between people's foreign policy goals and their more specific policy preferences. The strength of these connections is measured by standardized coefficients, analogous to standardized regression coefficients in multiple regression. That is, the coefficients are the estimated effect on a dependent variable, measured in 13

16 standard deviations, of a one standard deviation change in the independent or causal variable, controlled for the effects of the other independent variables. People who place a high value on military security goals are strong supporters of military intervention (.75) and military assistance policies (1.26). Such individuals are strong opponents (-.80) of policies designed to enhance US/Soviet cooperation, including arms control. In contrast, people who value economic security oppose military intervention (-.19), military assistance (-.73), and economic aid policies (-.26), and they support protectionist policies (.12). We observe, then, an interesting pattern in evidence. Ordinary citizens who support the general value of military security tend also to place a high value on economic security. But when each value is simultaneously controlled for the other, defenders of military security and defenders of economic security come down on different sides on the particular policies of military intervention and military assistance: military security supporters defend military intervention and assistance policies; economic security supporters oppose these same policies. Finally, we have already observed that ordinary people who support human rights goals tend to see no conflict between this value and support for military and economic security. But controlled for their security values, supporters of human rights are strong defenders of US/Soviet cooperation policies (.61) as well as economic assistance programs for third world nations (.77). The only policy that is not well-explained by the general goals is protectionism, which we predicted to be positively related to economic security values. The relationship is in the predicted direction (citizens who value economic security tend to support protectionist policies), but the association is comparatively small (.12) The weakness of this cognitive relationship may reflect 14

17 in part measurement error. This policy is measured by only one item, and it therefore has more random measurement error than the other latent policy positions. The 1986 survey includes no domestic policy alternatives that people may see as essential to economic security, and it has no policy items on energy independence, strategic minerals, or access to foreign markets. 5 Item sampling bias aside, however, American foreign policy in the post-war period has been defined most consistently in terms of military security, great power conflict, and human rights rather than by economic security concerns. Only since the Oil Embargo of 1973 and the emergence of the American trade deficit has American economic security seemed threatened by international events. It may be the case that by 1986 the mass public had not yet developed cognitive schema for considering policies that would advance economic security goals. Neither did such policies emerge as decisive issues in the 1988 presidential campaign, the attempt by Congressman Gephardt to exploit the issue in the Democratic nomination contest notwithstanding. 5 There is one item on grain sales to the Soviet Union, which is incorporated into the U.S.- Soviet cooperation measure, and one item on economic sanctions against South Africa, which is not included in this analysis. 15

18 People appear to see clear links between their policy preferences and their judgments of the Reagan Administration's performance in foreign affairs. The strongest endorsement of his performance comes from supporters of military assistance (.25) as well as from conservatives and Republicans (.61). In 1986, the Administration was judged unfavorably by defenders of U.S./Soviet cooperation (-.13). 6 In sum, Figure 2 offers persuasive evidence that people reason their way to judgments of presidential performance in foreign affairs by linking general goals to more specific policy positions. In covariance structure analyses, the goodness of fit of the theory is judged by summary measures that compare the observed correlation matrix to a matrix that is predicted by the hypothesized theory, that is, by both the measurement and the structural equation models. In Figure 2 the overall coefficient of determination for the structural equations is.97 and the overall goodness of fit index, adjusted for losses of degrees of freedom, is.92. These summary measures indicate that the overall model is well-specified and support the theory that there is a hierarchical cognitive structure by which the mass public considers foreign policy issues. The Leadership Sample 6 Two of the policy-performance paths have signs opposite to our expectations. Supporters of military intervention policies tend to be mildly critical of the Reagan Administration, a indication perhaps that they favored even tougher policies. And, protectionists tend to support Reagan in spite of his free trade positions. Both correlations, however, are relatively small, and there is a risk in over-interpreting them. 16

19 The elite sample include individuals in positions of leadership in government, academia (university presidents and faculty teaching foreign affairs), business, labor, the media, and religious institutions. The sampling frame includes members of the Reagan Administration as well as the House and the Senate. However, these 70 government officials were not asked the important questions on party identification or general liberal-conservative self-identification. Therefore, they are excluded from the analysis, leaving a sample of 273 opinion leaders. Differences between the leadership and the mass sample begin with the pattern of relationships among the general foreign policy goals. In the mass sample the three goals all display moderate to high positive intercorrelations, suggesting that these goals themselves derive from a higher order dimension such as nationalism-internationalism. In contrast, the elite respondents have a much more differentiated set of goals. 7 In Figure 3 the correlation between support for military security and economic security is only.29, compared to.70 in the mass sample. Even more striking are the relationships of human rights values to the other goals in the leadership sample. Leadership support for human rights is negatively correlated with support for military security (-.30) in contrast to the positive correlation in the mass sample, and the correlation between human rights and economic security beliefs falls to.18. For the elites human rights is a partisan and ideologically grounded value. Conservatives and Republicans consistently 7 A recurring issue in the study of ideologies is whether sophisticated belief systems are simple or complex. Whereas Stimson (1974) finds that cognitively sophisticated people tend to have a simpler, more integrated belief system, our evidence is that the elite sample has the more complex and differentiated beliefs on foreign policy values. 17

20 rank human rights goals on a lower scale of importance than liberals and Democrats, as the strong, negative correlation (-.66) between political leanings and human rights beliefs reveals. FIGURE 3 GOES ABOUT HERE Examining the relationships between the leaders goals and their policy preferences, we find that supporters of military security goals strongly endorse military intervention and military assistance policies (both associations are about.8), and they vigorously oppose US/Soviet cooperation (-.58). In contrast, the opinion leaders who are most concerned with U.S. economic security are apparently concerned about the opportunity costs of U.S. military commitments. Supporters of economic security goals tend to oppose military assistance policies (-.09) and military intervention policies (-.26). Unlike their counterparts in the cross-section sample, the economic security supporters stand beside the Reagan administration as opponents of protectionist trade policies as well(-.47), perhaps because they were more aware of Reagan's free trade positions than ordinary citizens were. The influence of policy positions and political leanings on judgments of the Reagan Administration displays the same pattern in the elite as in the mass sample. Reagan's strongest defenders are proponents of military assistance policies (.30). His detractors are concentrated among defenders of US/Soviet cooperation (-.35). Political leanings are strongly related to judgments of Reagan among elites (.45), just as they also are among ordinary citizens. 18

21 The leadership and the mass samples are also similar in the following respect: The cognitive links between goals and policies appear stronger than the links between policies and judgments of the Reagan administration. This pattern results in part from the fact that one dependent variable, judgment of presidential performance, is explained by five policy variables as well as political leanings. Each of the paths is an estimate of an effect of the causal variable, controlled for the others. Not all paths can be expected to be strong, therefore. The bivariate correlations between the policy variables and judgments of Reagan's performance are reasonably high, nonetheless. For example, in the mass sample, the correlations between performance judgments and the military intervention, military assistance, and U.S.-Soviet cooperation policy variables are.33,.56, and -.48, respectively. The comparable bivariate correlations for the leadership sample are.42,.80, and Overall, Figure 3 is strong evidence that U.S. leaders employ a hierarchical model of political reasoning. The coefficient of determination for the structural equations is.98, and the adjusted goodness of fit for the measurement and structural effects models is.91. The hierarchical model seems as well-specified for leaders as for ordinary citizens. 19

22 Summary and Conclusions This analysis supports the theory that there is a hierarchical structure to the belief systems of both the general public and opinion leaders. People do appear to behave as cognitive misers. Even if they have limited direct experience with international affairs, they can identify general goals for U.S. foreign policy, and they can link these goals to more specific policies. Similarly, people also seem to base their judgments of Reagan's performance in foreign affairs on the policy positions they hold. A social cognition theory of belief systems is quite consistent with the evidence of this analysis. The comparison of the cross-section and leadership samples indicates that the belief systems of ordinary citizens are as hierarchically structured as the beliefs of leaders. One major difference between leaders and non-leaders is that leaders more clearly distinguish human rights values from the goals of protecting American economic and military security. They appear less inclined to believe that American interests are consonant with the interests of other nations. A second important difference is that elites differentiate between the goals of military and economic security more sharply than do ordinary citizens. Undoubtedly, most Americans have a favorable image of the U.S. as a supporter of human rights. The public tends to accept a faith that U.S. military assistance and intervention policies are consistent with principles of advancing human rights and democracy abroad. The leadership sample is perhaps more realistic and certainly less trusting. It sees the national goals of economic and military security to be quite distinct from human rights. 20

23 This difference in conception of human rights in the two samples is also linked to partisanship and ideology. In the cross-section sample, Democrats and liberals were only slightly more likely to support human rights than Republicans and conservatives. In the leadership sample, the difference was quite marked, as reflected in the strong, negative correlation between the measures of human rights values and political leanings. In sum, the evidence is that ordinary citizens and opinion leaders alike reason about the foreign world in the manner predicted by a hierarchical social cognition model, a finding that reinforces similar conclusions by Wittkopf (1990) and Chittick, Billingsley, and Travis (1990). Moreover, the LISREL models also show that associational patterns in people's belief systems are not idiosyncratic, but rather are reasonably broadly shared. This sharing of policy bases for judgments of presidential performance is a necessary if not a sufficient condition for the existence of democratic accountability in the public's response to a government's policy choices. This shared bases for judgments is undoubtedly connected as cause and effect to the rising importance of foreign policy issues in contemporary U.S. elections. 21

24 REFERENCES Aldrich, John H., John L. Sullivan, and Eugene Borgida Waltzing Before a Blind Audience: The Anomaly of Foreign Affairs and Issue Voting. American Political Science Review. 83: Asher, Herbert B Presidential Elections and American Politics: Voters, Candidates, and Campaigns Since rd ed. Homewood, IL: Dorsey Press. Boyd, Richard Electoral Change and the Floating Voter: The Reagan Elections. Political Behavior. 8: Popular Control of Public Policy: A Normal Vote Analysis of the 1968 Election. American Political Science Review. 66: , with Paul R. Mencher, Philip J. Paseltiner, Ezra Paul, and Alexander S. Vajda The 1984 Election as Anthony Downs and Stanley Kelley Might Interpret It. Political Behavior. 10: Caloss, Dario Jr The Foreign Policy Belief Systems of Americans: National Idealism and National Self-Interest as Central Value Systems Among the American Public. Paper presented to Joint Annual Meeting of the British International Studies Association and the International Studies Association, London, March 18-April 1. Campbell, Angus, Philip E. Converse, Warren E. Miller, and Donald E. Stokes. The American Voter New York: Wiley. Carmines, Edward G. and James A. Stimson Issue Evolution: Race and the Transformation of American Politics. Princeton: Princeton University Press. 22

25 Chittick, William O., Keith R. Billingsley, and Rick Travis American Foreign Policy Beliefs: The Perils of Flatland. Paper presented to Joint Annual Meeting of the British International Studies Association and the International Studies Association, London, March 18-April 1. Chittick, William O., Keith R. Billingsley, and Rick Travis Persistence and Change in Elite and Mass Attitudes Toward U.S. Foreign Policy. Political Psychology. 11: Conover, Pamela Johnston and Stanley Feldman How People Organize the Political World: A Schematic Model. American Journal of Political Science. 28: Downs, Anthony An Economic Theory of Democracy. New York: Harper and Row. Enelow, James M., Melvin J. Hinich, and Nancy R. Mendell An Empirical Evaluation of Alternative Spatial Models of Elections. Journal of Politics. 48: Fiske, Susan T. and Shelley E. Taylor Social Cognition. Reading, MA: Addison- Wesley. Gant, Michael M. and Dwight F. Davis Mental Economy and Voter Rationality: The Informed Citizen Problem in Voting Research. Journal of Politics. 46: Graham, Thomas W The Pattern and Importance of Public Knowledge in the Nuclear Age. Journal of Conflict Resolution. 32: Hamill, Ruth C., Milton Lodge, and Frederick Blake The Breadth, Depth, and Utility of Class, Partisan, and Ideological Schemata. American Journal of Political Science. 29:

26 Holsti, Ole R. and James N. Rosenau American Leadership in World Affairs: Vietnam and the Breakdown of Consensus. Boston: Allen and Unwin The Structure of Foreign Policy Attitudes among American Leaders. Journal of Politics. 52: Hurwitz, Jon and Mark A. Peffley. 1987a. How Are Foreign Policy Attitudes Structured? A Hierarchical Model. American Political Science Review. 81: b. The Means and Ends of Foreign Policy as Determinants of Presidential Support. American Journal of Political Science. 31: Alternative Modes of Political Reasoning in Foreign Affairs. Presented at the Annual Meeting of the Midwest Political Science Association, April 14-16, Chicago Public Images of the Soviet Union and Its Leader: The Impact of Foreign Policy Attitudes. Journal of Politics. 52:3-28. Jöreskog, Karl G. and Dag Sörbom LISREL VI: Analysis of Linear Structural Relationships by Maximum Likelihood, Instrumental Variables, and Least Squares Methods. Mooresville, IN: Scientific Software, Inc. Iyengar, Shanto Television News and Citizens' Explanations of National Affairs. American Political Science Review. 81: Long, J. Scott. 1983a. Confirmatory Factor Analysis: A Preface to LISREL. Beverly Hills, CA: Sage b. Covariance Structure Models. Beverly Hills, CA: Sage. 24

27 Maggiotto, Michael A. and Eugene R. Wittkopf American Public Attitudes Toward Foreign Policy. International Studies Quarterly. 25: Nie, Norman H., Sidney Verba, and John R. Petrocik The Changing American Voter. Cambridge: Harvard University Press. Page, Benjamin I. and Robert Y. Shapiro Changes in Americans' Policy Preferences, Public Opinion Quarterly. 46: Peffley, Mark A. and Jon Hurwitz A Hierarchical Model of Attitude Constraint. American Journal of Political Science. 29: Pomper, Gerald M From Confusion to Clarity: Issues and American Voters, American Political Science Review. 66: RePass, David E Issue Salience and Party Choice. American Political Science Review. 65: Russett, Bruce Democracy, Public Opinion, and Nuclear Weapons. In Philip E. Tetlock, Jo L. Husband, Robert Jervis, Paul C. Stern, and Charles Tilly (eds.), Behavior, Society, and Nuclear War. Vol. 1. New York: Oxford University Press. Smith, Jeffrey A American Presidential Elections: Trust and the Rational Voter. New York: Praeger. Smith, Tom W America's Most Important Problem A Trend Analysis, Public Opinion Quarterly. 44: Stimson, James A Belief Systems: Constraint, Complexity, and the 1972 Elections. American Journal of Political Science. 19:

28 Stokes, Donald E Spatial Models of Party Competition. In Angus Campbell et al., Elections and the Political Order. New York: Wiley. Taylor, Shelley E. and Jennifer Crocker Schematic Bases of Social Information Processing. In Social Cognition: The Ontario Symposium, Vol 1, E. Tory Higgins, C. Peter Herman, and Mark P. Zanna, eds. Hisdale, NJ: Erlbaum. Wittkopf, Eugene R Elites and Masses: Another Look at Attitudes toward America's World Role. International Studies Quarterly. 31: Faces of Internationalism: Public Opinion and American Foreign Policy. Durham: Duke University Press. Wolfers, Arnold Discord and Collaboration. Baltimore: Johns Hopkins. 26

29 APPENDIX A: MEASURES OF LATENT VARIABLES Foreign Policy Attitudes Item Number Item Description Mass Leadership I. Goals 8 A. Economic Security Q21C "Reducing our trade deficit with foreign countries" Q18C "Keeping up the value of the dollar" Q21G Q18G "Protecting the jobs of American workers" Q21H Q18H "Protecting the interests of American business abroad" Q21J Q18J "Securing adequate supplies of energy" B. Military Security Q21A Q18A "Protecting weaker nations against foreign aggression" Q21F Q18F "Matching Soviet military power" Q21K Q18K "Defending our allies' security" Q21N Q18N "Containing Communism" C. Human Rights Q21B Q18B "Promoting and defending human rights in other countries" Q21D Q18D "Strengthening the United Nations" Q21E Q18E "Combatting world hunger" Q21I Q18I "Helping to bring a democratic form of government to other nations" Q21L Q18L "Worldwide arms control" Q21M Q18M "Helping to improve the standard of living of less developed countries" II. Political Leanings Q45A,B,C Q43A,B,C Party Identification Index, leaners categorized with weak partisans. Q46 Q44 Ideological self-placement, very conservative to very liberal. 8 As Appendix B shows, 4 of the goal items were specified as measures of a second latent factor in addition to the one listed below. 27

30 III. Policy Preferences A. Military Intervention Q20_ of 11 items on "circumstances that might justify using U.S. troops in other parts of the world..." Q15_1-7 5 items of the above 10 B. Military Assistance Q5D "Military aid to other nations..." Expand, cut back Q10A Q6 "...favor or oppose giving military aid to other nations..." Q10B "...favor or oppose selling military equipment to other nations..." Q11 Q7 "...give economic and military aid to rebel groups fighting their communist-supported governments..." Q22C Q19C "The U.S. may have to support some military dictators because they are friendly toward us..." Q36 Q35 "...do you feel the CIA should or should not work secretly to try to weaken or overthrow governments..." C. U.S.-Soviet Cooperation Q22A Q19A "The U.S. has not tried hard enough to make an agreement with the Soviet Union to reduce the number of nuclear weapons on both sides." Q30_2,3 Q30_2,3 2 items on relations between Soviet Union and U.S. Q26 Q24 "The U.S. should stop building nuclear weapons even if the Soviet Union does not." Q27 Q25 "What about the idea of a mutual, verifiable freeze..." Q28 Q26 "From what you have heard about the proposed Strategic Defense Initiative or "Star Wars" system..." D. Economic Aid Q5E "Economic aid to other nations..." Expand, cut back Q9 Q5 "On the whole, do you favor or oppose giving economic aid to other nations for purposes of economic development and technical Assistance..." Q41 Q41 "Should the U.S. give other countries foreign aid to help their farmers grow different crops and not depend on illegal drugs for their income"... E. Protectionism Q43 Q38 "...would you say you sympathize more with those who want to eliminate tariffs...? IV. Judgments of Reagan's and U.S. Government's foreign policies Q15_1-8 Q11_1-8 7 of 8 items on "Reagan Administration's handling" of problems. Q16_1-4 Q12_1-4 3 of 4 items on "actions of the U.S. government..." 28

31 APPENDIX B I. Standardized Loadings of the Observed Endogenous (Dependent) Items: Mass Sample* Military Military US/Soviet Intervention Assistance Cooperation Q20_1.659 Q5D.539 Q22A.530 Q20_2.779 Q10A.775 Q30_2.353 Q20_3.488 Q10B.595 Q30_3.306 Q20_4.780 Q Q Q20_5.786 Q22C.469 Q Q20_6.707 Q Q Q20_7.779 Q20_8.657 Q20_9.697 Q20_ Economic Aid Protectionism Reagan's Performance in Foreign Affairs Q5E.592 Q Q15_1.844 Q9.739 Q15_2.750 Q Q15_3.638 Q15_4.792 Q15_6.473 Q15_7.745 Q15_8.624 Q16_1.599 Q16_2.705 Q16_4.663 II. Standardized Loadings of the Observed Endogenous Items: Elite Sample* Military Military US/Soviet Intervention Assistance Cooperation Q15_1.586 Q6.573 Q19A.852 Q15_2.747 Q7.893 Q30_2.511 Q15_3.644 Q19C.709 Q30_3.584 Q15_4.349 Q Q Q15_7.790 Q Q Economic Aid Protectionism Reagan's Performance in Foreign Affairs Q5.810 Q Q11_1.846 Q Q11_2.598 Q11_3.747 Q11_4.852 Q11_6.409 Q11_7.818 Q11_8.615 Q12_1.818 Q12_

32 Q12_4.868 Foreign Policy Attitudes *Beside each questionnaire number is a factor loading. The loading is the correlation between the item and the latent dimension it measures. The latent dimension Protectionism is measured by only one item. Its loading is therefore fixed at 1.0, with a zero error variance. Each item loads on only one latent variable. 30

33 III. Standardized Loadings of the Observed Exogenous (Independent) Variables: Mass and Elite Samples Mass Sample Elite Sample Economic Military Human Political Economic Military Human Political Security Security Rights Leanings Security Security Rights Leanings Q21C.523 Q18C.429 Q21G.451 Q18G.587 Q21H Q18H Q21J.678 Q18J.774 Q21A Q18A Q21F.650 Q18F.679 Q21K Q18K Q21N.656 Q18N.742 Q21B.667 Q18B.482 Q21D.467 Q18D.667 Q21E.635 Q18E.760 Q21I Q18I Q21L.534 Q18L.790 Q21M.722 Q18M.558 Party ID.571 Party ID.893 Q Q *For theoretical reasons, items H, A, K, and I were permitted to load on more than one factor in both samples. For example, item I, "Helping to bring a democratic form of government to other nations," could be considered either a military security goal or a human rights goal, depending on the perspective of the individual respondent. 31